Dynamics, control and extremum seeking of the Rectisol process

Gasification based biorefineries have been studied in the past decade as part of a global effort to replace fossil fuels to produce energy and added value chemicals. An important part of these biorefineries is the acid gas removal units, that remove CO2 and H2S from the produced synthesis gas. One of the acid gas removal processes associated in these studies is Rectisol. Rectisol has been chosen since it's environmental friendly and requires a lower amount of operational and capital costs compared to its opponents.;To carry out a dynamic study of the process, as a first step, a steady-state simulation was carried out in Aspen Plus RTM. The steady-state behavior of the columns were studied and validated based on data found in the literature. Control valves were placed in all the necessary places. After sizing the equipment, such as seperation drums, valves and column sumps, the pressures were varified, so that the pressure at the inlet of each equipment corresponds to incoming stream. Later on the model was exported to Aspen plus dynamics and the effort of different inputs and disturbances on the outputs were studied.;Due to the fact that the composition of the gasified biomass varies, the composition and the amount of impurities in the gasification gas also varies This creates variations in the purities of the syngas and byproducts of acid gas removal units. In any chemical plant it is important to keep compositions of products as constant as possible so that we don't create perturbations in downstream units. To overcome these variations an adaptive extremum control scheme was implemented that optimizes a quadratic objective function of product compositions, while the relation between the objective function and its independent variables is unknown.;For the adaptive extremum seeking control to be effective, a responsive plantwide regulatory control structure is required. Absorption based gas cleaning processes like Rectisol all have a recycle flow of solvent. This recycle flow can always be problematic from a process control point of view. Thus a search was conducted amongst the conventional and advanced control techniques. Four potential control strategies were implemented and their performance was analyzed. These four strategies were Multiloop PI, Centralized Model Predictive Control (MPC), Decentralized MPC and Distributed MPC. The reason we have chosen MPC is that these controllers can systematically consider process variable interactions and input and output constraints in their control calculations. Among the four, distributed and centralized MPC were found to be most effective in terms of rejecting input flow disturbances and tracking setpoints. Keeping this fact in mind a multivariable extremum-seeking scheme was designed and implemented on these two types of controllers and their performance was studied for different dither signal frequencies. The results showed that at the proper frequency, both combination of optimization and control structures have identical behavior.;At the end the combination of adaptive extremum seeking and Distributed MPC was chosen as the optimizing and control structure for the studied Rectisol plant, since Distributed MPC is more fault tolerant and the control of the plant will not depend on a single control agent. In conclusion, Rectisol has been robustified to the composition and flowrate of the input and the plant is able to keep its product compositions as close as possible to the desired specifications.